def response(flow):
url = flow.request.url
orig_path = get_path(url, origdir)
orig_dir_path = os.path.dirname(orig_path)
if not os.path.exists(orig_dir_path):
os.makedirs(orig_dir_path)
with open(orig_path, 'wb') as fp:
fp.write(flow.response.content)
mod_path = get_path(url, moddir)
if os.path.exists(mod_path):
# rewrite response
with open(mod_path, 'rb') as fp:
flow.response.content = fp.read()
def poor_log(msg, *args):
from time import strftime
from lib import get_path
timestamp = strftime("%Y-%m-%d %H:%M:%S")
log_file = get_path('log')
with open(log_file, 'a+b', 0) as f:
format = "[%s]" % timestamp + msg + "\n"
f.write(format % args)
def edit(url):
origpath = get_path(url, origdir)
modpath = get_path(url, moddir)
editor = os.getenv('EDITOR', 'vim')
cp_ts = None
if not os.path.exists(modpath) and os.path.exists(origpath):
modpath_dir = os.path.dirname(modpath)
if not os.path.exists(modpath_dir):
os.makedirs(modpath_dir)
shutil.copyfile(origpath, modpath)
cp_ts = get_ts(modpath)
ret = subprocess.call([editor, modpath])
# delete if no changes on copied file
if cp_ts and cp_ts == get_ts(modpath):
logger.info('no changes on {}'.format(modpath))
os.remove(modpath)
def __init__(self, proj_path, cfg, network = None):
self._cfg = cfg
self._root_path = proj_path
self.net = network
# self.output_dir = output_dir
pretrained_model_path = os.path.join(self._root_path,
cfg.COMMON.DATA_PATH, cfg.TRAIN.PRETRAIN)
pretrained_model = os.listdir(pretrained_model_path)
assert len(pretrained_model) == 1, 'pretrain model should be one'
self._pretrain = os.path.join(pretrained_model_path, pretrained_model[0])
self._ckpt_path = get_path(os.path.join(self._root_path, cfg.COMMON.DATA_PATH, cfg.TRAIN.CKPT))
self._restore = cfg.TRAIN.RESTORE
self._max_iter = cfg.TRAIN.MAX_ITER
if __name__ == '__main__':
# print(os.getcwd())
cfg = load_config()
print('Using config:')
# pprint.pprint(cfg)
"""
@params
use_cache 是否从重新进行data_process过程，一般dataset/for_train文件发生变化需要进行
"""
roidb = get_training_roidb(cfg) # 返回roidb roidb就是我们需要的对象实例
# output_dir = get_path('dataset/output')
# log_dir = get_path('dataset/log')
checkpoints_dir = get_path(cfg.COMMON.CKPT)
# print('Output will be saved to {:s}'.format(output_dir))
# print('Logs will be saved to {:s}'.format(log_dir))
# """
# @params
# network ctpn_network 实例
# roidb roi 列表
# output_dir tensorflow输出的 绝对路径 要拼接本地机器所在的运行目录
# log_dir 日志输出 绝对路径
# max_iter 训练轮数
# pretrain_model 预训练VGG16模型 绝对路径
# restore bool值 是否从checkpoints断点开始恢复上次中断的训练
# """
vgg16_net_param = '../dataset/pretrain/VGG_imagenet.npy'
total_rows += sheet.max_row
keys = {}
for j, row in enumerate(sheet.rows):
if j > 0:
break
for k, cell in enumerate(row): # Проверяем, чтобы был СНИЛС и Код
if cell.value in IN_SNILS:
keys[IN_SNILS[0]] = k
if len(keys) < 1:
print('В файле "' + sys.argv[i + 1] +
'" отсутствует колонка со СНИЛС')
time.sleep(3)
sys.exit()
sheets_keys.append(keys)
path = get_path(sys.argv[1])
print('\n' + datetime.now().strftime("%H:%M:%S") +
' Начинаем преобразование и нарезку xlsx файлов \n')
k = 1 # Счетчик строк в csv
file_number = 1
cl_csvs = []
for i, sheet in enumerate(
sheets): # Загружаем все xlsx файлы по мере сохранения в БД
for j, row in enumerate(sheet.rows): # Теперь строки
if j == 0:
continue
cl_csv = {}
if lenl(row[keys[IN_SNILS[0]]].value
) < 12: # and l(row[keys[IN_SNILS[0]]].value) > 100:
from ctpn.train_net import train_net
from lib.load_config import load_config
from data_process.roidb import get_training_roidb
from lib import get_path
if __name__ == '__main__':
cfg = load_config()
print('Using config:')
pprint.pprint(cfg)
"""
@params
use_cache 是否从重新进行data_process过程，一般dataset/for_train文件发生变化需要进行
"""
roidb = get_training_roidb(cfg) # 返回roidb roidb就是我们需要的对象实例
output_dir = get_path('dataset/output')
log_dir = get_path('dataset/log')
checkpoints_dir = get_path(cfg.COMMON.CKPT)
print('Output will be saved to {:s}'.format(output_dir))
print('Logs will be saved to {:s}'.format(log_dir))
# """
# @params
# network ctpn_network 实例
# roidb roi 列表
# output_dir tensorflow输出的 绝对路径 要拼接本地机器所在的运行目录
# log_dir 日志输出 绝对路径
# max_iter 训练轮数
# pretrain_model 预训练VGG16模型 绝对路径